One example is with databases. In the olden days, if you needed performance, you bought a big Oracle server, and that was that. If you needed more performance, you just bought a bigger one. Without revisiting the database from first principles, we'd still be using acid-compliant relational SQL databases for everything. By starting from first principles and asking what the database software is for, and considering a changed hardware landscape aka the cloud, among other factors, we now have other types of databases which are more appropriate for different kinds of workloads. Key-Value stores, document stores, Wide column stores, Graph databases, time-series databases, NewSQL databases, vector databases; all of those are an example of first-principles rethinking of the database, and being willing to throw out the rules and ACID compliance in favor of what's actually necessary for a given workload instead of continuing on with existing conventions.